Models: TypeError: export_inference_graph() got an unexpected keyword argument 'use_side_inputs'

[chuasonglin]

I tried following Edje Electronics's tutorial on for training an object detector.
https://www.youtube.com/watch?v=Rgpfk6eYxJA

I have managed to train the model already. But I have problems with exporting inference graph.
TypeError: export_inference_graph() got an unexpected keyword argument 'use_side_inputs'

tensorflow version: 1.15

[ayjomoluabi]

I have the exact same issue, while using tensorflow 1.14.0

TypeError: export_inference_graph() got an unexpected keyword argument 'use_side_inputs'

Any help would be appreciated.

[chuasonglin]

I have the exact same issue, while using tensorflow 1.14.0

TypeError: export_inference_graph() got an unexpected keyword argument 'use_side_inputs'

Any help would be appreciated.

I somehow managed to get passed it. I'm thinking that the files we have are not compatible anymore. Some are updated while some are not. So I downloaded a very old version of the export_inference_graph.py from here:

https://github.com/datitran/object_detector_app/blob/master/object_detection/g3doc/exporting_models.md

And here are the codes to type instead:

python export_inference_graph.py --input_type image_tensor --pipeline_config_path training/faster_rcnn_inception_v2_pets.config --checkpoint_path training/model.ckpt-7359 --inference_graph_path output_inference_graph.pb

hope it helps you too

[ayjomoluabi]

I have the exact same issue, while using tensorflow 1.14.0
TypeError: export_inference_graph() got an unexpected keyword argument 'use_side_inputs'
Any help would be appreciated.

I somehow managed to get passed it. I'm thinking that the files we have are not compatible anymore. Some are updated while some are not. So I downloaded a very old version of the export_inference_graph.py from here:

https://github.com/datitran/object_detector_app/blob/master/object_detection/g3doc/exporting_models.md

And here are the codes to type instead:

python export_inference_graph.py --input_type image_tensor --pipeline_config_path training/faster_rcnn_inception_v2_pets.config --checkpoint_path training/model.ckpt-7359 --inference_graph_path output_inference_graph.pb

hope it helps you too

Thanks mehn, this worked! However I am concerned how efficient the process was.
Fingers crossed.

[pkulzc]

We kept updating the whole repo and it looks like your files are not sync'ed to the same stamp/commit.
Instead of downloading an old copy, you should sync your repo to latest :)

[joe983]

Following these instructions on colab

https://towardsdatascience.com/detailed-tutorial-build-your-custom-real-time-object-detector-5ade1017fd2d

At this step:
!python /content/drive/'My Drive'/object_detection/models/research/object_detection/export_inference_graph.py \ --input_type=image_tensor \ --pipeline_config_path=/content/drive/'My Drive'/object_detection/models/research/object_detection/samples/configs/ssd_mobilenet_v2_coco.config \ --output_directory={output_directory} \ --trained_checkpoint_prefix={last_model_path}

I had same issue (TypeError: export_inference_graph() got an unexpected keyword argument 'use_side_inputs')

Tried with updated version of export_inference_graph.py, as suggested, and it didn't work (gave same error).

Tried with old version of export_inference_graph.py (as mentioned above), and it gave this error:

AssertionError: Inference graph path missing

Added 'inference_graph_path' line to code:

!python /content/drive/'My Drive'/object_detection/models/research/object_detection/export_inference_graph.py \ --input_type=image_tensor \ --pipeline_config_path=/content/drive/'My Drive'/object_detection/models/research/object_detection/samples/configs/ssd_mobilenet_v2_coco.config \ --inference_graph_path={output_directory} \ --output_directory={output_directory} \ --trained_checkpoint_prefix={last_model_path}

and got following error:

ValueError: ('%s is not decorated with @add_arg_scope', ('nets.mobilenet.mobilenet', 'depth_multiplier'))

Any ideas where I'm going wrong?

Sorry I just realised this question shouldn't be on here.
And I fixed it.
Should 've added 'inference_graph_path' line to code as:

--inference_graph_path output_inference_graph.pb \

[nachiketmore99]

Deleting the folder where your inference graph is saved (poker_inference_graph in your case) should work, it worked for me

[cheadrian]

Delete the variables and flags that refers to side_input should get rid of that error.
Can't guarantee if the network will still work as expected after this.

Here's the code:
https://gist.github.com/cheadrian/f02363499bdfa826b1c311c224c89215